Research On The Schduling Algorithm For Network-on-chip

In previous SoC design, bus structure and point-to-point structure were two structures which were often used for connecting with peocessors. However, both of them gradually revealed some shortcomings as the increasing number of processors in system and led to a series of problems, such as poor scalability and reusability of resource in SoC designs. The emergence of NoC has successfully helped solve those problems existed in current SoC design. Noc has various research aspects including microstructure design of hardware, data communication protocol, data flow control technology as well as task scheduling. Among them, recently, task scheduling has been the focus in academia.In the last decade, task scheduling peoblem has been extensively explored. A good scheduling algorithm plays an important role in improving the performance of NoC system. One dominant merit of a good algorithm is that it enables to make full use of on-chip resources and reasonably arrange the executing order of tasks. This could realize parallel computation and minimize the scheduling length with improving the efficiency of system. Another benefit of a great scheduling algorithm, which is an important issue on NoC, is to minize the data transmission distance among tasks as far as possible in order to reduce the routing energy. What’s more, an excellent scheduling algorithm can optimize multiple performance metrics, such as workload balance and thermal management.This article focuses on the task scheduling problem on NoC, and mainly aims at the DAG task scheduling. The main content of the paper is organized as follows:1. Analyze the heuristic scheduling algorithms including Genetic Algorithm, Discrete Particle Swarm Optimization and Simulated Annealing. Besides, compare the simulation results of these algorithms.2. Propose ADIA based on improvement of the artificial immune system and introducing NS technology. Verify the validity of ADIA in improving system scheduling performance with the simulation results.3. Introduce Data Envelopment Analysis in the research of tash scheduling on NoC. Construct multi-output Decision Making Unit by extracing four common metrics form the task scheduling problem and propose the BCC referred-time evaluation for the scheduling solution. Finally, put forward BCC-DPSO by combining the evaluation method and discrete particle swarm optimization. As a result of the simulation, it shows that our proposal always derived the best efficiency among all the five algorithms.